DAAD: Dynamic Analysis and Adaptive Discriminator for Fake News Detection

Xinqi Su,Yawen Cui,Ajian Liu,Xun Lin,Yuhao Wang,Haochen Liang,Wenhui Li,Zitong Yu

2024-08-20

Abstract:In current web environment, fake news spreads rapidly across online social networks, posing serious threats to society. Existing multimodal fake news detection (MFND) methods can be classified into knowledge-based and semantic-based approaches. However, these methods are overly dependent on human expertise and feedback, lacking flexibility. To address this challenge, we propose a Dynamic Analysis and Adaptive Discriminator (DAAD) approach for fake news detection. For knowledge-based methods, we introduce the Monte Carlo Tree Search (MCTS) algorithm to leverage the self-reflective capabilities of large language models (LLMs) for prompt optimization, providing richer, domain-specific details and guidance to the LLMs, while enabling more flexible integration of LLM comment on news content. For semantic-based methods, we define four typical deceit patterns: emotional exaggeration, logical inconsistency, image manipulation, and semantic inconsistency, to reveal the mechanisms behind fake news creation. To detect these patterns, we carefully design four discriminators and expand them in depth and breadth, using the soft-routing mechanism to explore optimal detection models. Experimental results on three real-world datasets demonstrate the superiority of our approach. The code will be available at: <a class="link-external link-https" href="https://github.com/SuXinqi/DAAD" rel="external noopener nofollow">this https URL</a>.

Artificial Intelligence,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper attempts to address the issue of the rapid spread of fake news on social media in the current network environment, which poses a serious threat to society. Existing multimodal fake news detection (MFND) methods are mainly divided into knowledge-based methods and semantic-based methods. However, these methods overly rely on human expertise and feedback, lacking flexibility. To this end, the paper proposes a Dynamic Analysis and Adaptive Discriminator (DAAD) method to detect fake news.

Specifically, the paper improves on the following two main issues:

1. **Knowledge-based methods**:
- Existing knowledge-based methods overly rely on static knowledge bases, making it difficult to cope with increasingly complex fake news.
- The paper introduces the Monte Carlo Tree Search (MCTS) algorithm, utilizing the self-reflection capability and error feedback of large language models (LLMs) to optimize prompts, thereby providing richer, domain-specific details and guidance, enabling LLMs to more flexibly comment on news content.

2. **Semantic-based methods**:
- Although existing semantic-based methods can capture some characteristics of fake news, they still have the problem of manually constructing models and relying on human knowledge, resulting in detection models that are not flexible and comprehensive enough.
- The paper defines four typical deception patterns: emotional exaggeration, logical inconsistency, image manipulation, and semantic inconsistency, and designs four discriminators to detect these patterns. Through a soft routing mechanism, the depth and breadth of these discriminators are expanded to explore the optimal detection model.

Overall, the paper aims to improve the accuracy and flexibility of fake news detection through the Dynamic Analysis and Adaptive Discriminator method, reducing reliance on human expertise.